Originally Posted by CL3P20
There is NO "max safe votage".. only the max voltage you decide to pump through the poor CPU. Whatever the manufacturer rating is, we all know from previous experience, that the 'limits' set by Intel are easily and relatively safely exceeded, provided you have the cooling potential to keep your CPU below 60c or so.
Also, you should always go by the voltage as set in BIOS..as your "Real Vcore" ...the amount of voltage seen in Windows is after "Vcore Droop".
*Max I will run on a 45nm CPU is ~1.6v on water for benching and 1.55v for daily use..~1.7v benching for 65nm and 1.65v for daily use.
Don't get me started on the affects of errata. Even if you cooled the processor sufficiently the immense amount of electricity flowing through the transistors in the processor creates magnetic fields strong enough to cause EMI (electro-magnetic-interference) between the nano sized gaps between transistors, causing processor instablility/crashes. That's why an i7 can only reach 5.5 GHZ even with liquid nitrogen keeping the processor at below freezing temperature (not to mention extreme measures for chipset and RAM cooling), but imagine how long all those components would really last even if you were to keep the cooling constant...
My definition of safe voltage is keeping the CPU alive and stable for eternity/next CPU upgrade. If I get another 1.2 GHZ staying within in the limits, then I'm satisfied. Intel set standards, but I honestly don't remember Intel being so vocal about "frying CPU/RAM" in their first announcement of a new processor. When the first 65nm came out they never publicly announced "DON'T VOLT THIS UP". You think Intel forced all Motherboard and RAM manufacturers to bump their vdimm down to 1.5 for ****S and GIGGLES? I think not.
i7's are new so we haven't had time to really see. You can gamble, but I'd rather play it safe until I buy my next CPU, if that means sacrificing a measley 200-300 MHZ for 100% life and stability so be it.
I guess I'm talking more about the RAM than the CPU but they're connected so it viable to say the RAM can take the CPU with it.
Life of DDR3 at Tested Voltages sufficiently cooled:
1.90v vdimm / 1.1 vtt = dead after 10 hours
1.65v vdimm / 1.20 vtt = ok forever
1.80v vdimm / 1.30 vtt = ok for 4 weeks
1.95v vdimm / 1.40 vtt = ok for 2 weeks
2.05v vdimm / 1.45 vtt = ok for 2 weeks
2.30v vdimm / 1.35 vtt = ok for 1 week (cold)
I didn't see the actual calculations but I'm guessing it's an exponential effect. At 1.7v vdimm maybe its like a year and 1.75v vdimm is 6 months, who knows. If your RAM can't support the OC at 1.65v you're gambling. You also aren't thinking that the memory controller are now integrated in the i7 which means the electricity is bound directly with the vdimm this changes things A LOT.Edited by Cyberpunx - 12/27/08 at 12:51pm